Three-dimensional Reconstruction And Attitude Estimation Base On High Resolution Image Sequences

ISAR(Inverse synthetic aperture radar)is a high resolution imaging device,which can detect,recognize and monitor small targets,such as aircraft,missiles and satellites all day long,actively,real-time and all day.How to obtain the three-dimensional information of the observed target from the two-dimensional information of high resolution imaging supports the decision making based on the data obtained by the inverse synthetic aperture radar(SAR).In this paper,three dimensional reconstruction of key targets based on highresolution image sequences is completed,and the attitude estimation of 3D focused targets is further accomplished.This paper first introduces the background of the research and the significance of the research.Through the analysis of the current research on corner detection,3D reconstruction and attitude estimation,it lays the foundation for the research in this paper.Considering the inherent characteristics of imaging equipment and ISAR imaging mode,high resolution image preprocessing operation has been targeted.The particularity of radar imaging principle has led to the particularity of image noise.In this paper,median filtering is used to suppress speckle noise in images.Long distance observation and imaging will inevitably lead to partial blurring,breakage and adhesion of images.In order to obtain better two-dimensional image information,this paper uses the principle of mathematical morphology to carry out "fill in the surplus" operation.Next,the SIFT algorithm is used to extract and match the feature points of the image.After the extraction and matching of image features,robust features in ISAR sequence images are obtained.In this paper,the main main factor decomposition method is used to establish the 3D target reconstruction based on the feature of sequence image.In the attitude estimation part,the three-dimensional attitude information database of the key target is first set up in this paper.Each image contains three dimensional attitude information when the image is taken.Defining similarity and setting threshold to estimate the pose of a particular target at a given time.Matching the image of the target at that time with the image in the database,and estimating the 3D pose of the key target according to the image in the database.In this paper,the simulation experiment of three dimensional attitude estimation based on the image sequence is completed,and a twolayer estimation algorithm is proposed to achieve fast,ready and high efficiency estimation.